I would be willing to bet the 60Wh quote is in regards to the monitoring system and pumps needed to move the nutrient solution, not to move the towers.
"You aren't going to get one through hydroponic farming. Hydroponic tomatoes look beautiful and big but they are dense and flavorless."
That would be the cultivar chosen for mechanical picking and shipment, plus a combination of poor nutrient choice. Get an heirloom cultivar and you will notice a world of difference in taste, using the same nutrients. Using more diverse stuff like SEA-90 (supplementing for the nitrogen and potassium [I think, don't have that chart handy right this second]) will show even better results.
Here's a soil mother (two months old) tomato and basil versus a hydroponics tomato and basil clone (three weeks old.) Look at the difference.
The basil alone was DRIPPING with pungent oils, and much larger leaves. I had pesto that had my local Italian neighbors raving for WEEKS.
If you have a good room with barriers, not much pest control.
That's not vertical farming, though. That's rotary. Vertical implies large stacks of trays or channels. Most weed growers use 'vertical' for the vertically-oriented NFT systems, but that's not true vertical.
They taste fine. I've done a grafting of potato rootstock + Medusa peppers. Potatoes weren't spicy, peppers weren't starchy and properly spicy (though it looked really strange having a part potato vine loaded with clusters of peppers.)
"I hope they consulted with the experts in the field, the weed growers."
Those aren't experts. Expert botanists use photon flux density as their measurement, not lumens. Any electrician will tell you how bullshit the term "grams per watt" is when used how the weed growers use it. Any photobiologist will know that going past ~3,000 umol (50% more photon flux than the sun puts out per meter at sea level mid-summer at noon) is a BAD idea for most plants, including cannabis (yet idiot growers want those white, photosynthetic-bleached 'frosty' nugs.)
The fact that you refer to these people as 'experts' is laughable and shameful to those of us that have REAL expertise in this field, across more crops than just petty weed and tomatoes and lettuce.
Shit I can get you 150+ lumens per watt but that isn't what matters, photon flux density per wavelength is what matters.
And I can custom-build you a panel at pretty much near-cost. I've got the capability to manufacture units so small and so powerful that Intel would be scratching their heads wondering how the fuck I keep them cool.
500w in 30mm x 30mm? Easily done for the raw array. Copper-backed, attached to a large aluminum radiator, no problem. Power supply that's reliable? You want that internal or external?
It's only expensive because you're looking at scamming resellers of Chinese stuff. Go check out Hydrogrowled, then look at the patents they claim, only to realize that's what every other Chinese manufacturer on the planet has already claimed and made, well before those patents were even filed.
"Everyone keeps mentioning LED but LED is insanely expensive lighting"
No it's not. You can order direct and custom from the Chinese, 90w panels to your specs for like $75 each - the REAL issue is whether or not you know what you WANT in the panel. That might cost you extra (Like aluminium thermal PCB instead of PolyPCB+Heat Plate attachment) and such.
Cut out the middleman and go direct to the REAL manufacturer.
Let me guess, you looked at the scammer site hydrogrowled.com, didn't you?
"Maximum theoretical overall efficiency of LED lighting is only 43.9 % (http://en.wikipedia.org/wiki/Luminous_efficacy)."
Outdated information. Did you forget the LED in the lab that output more light energy than it consumed, by virtue of the surrounding thermal energy? That was a story right here on slashdot not even a month or two ago.
Also, 299 lumens per watt for LED? Cree's already hit over 220 in the labs. We have new topology and substrate manufacturing that eliminates many of the inefficiencies of LED (backscatter, re-absorption of emitted photons,) that have yet to be implemented in common process.
At the rate I've watched LED tech progress, in maybe three years we'll hit that barrier and break it, we have power supplies with power factors of 0.98, the only inefficiency is A. source power and B. LED
"I understand plant photosynthesis can't use the IR, a narrow green band, and UV portions of sunlight"
This is incorrect. IR has its uses in triggering flowering hormones plus day/night cycles, plus it has an additional photosynthetic effect. Green is used - its quantum yield is higher than blue or red, but ONLY when the photon flux density gets roughly over ~400 umol (which most vegetative/non-fruiting crops do well within 150-300 umol) and UV has photosynthetic and carotenoid reactions - and for those growing cannabis, UVB light specifically has a boosting effect upon THC production in plant tissues and UVA has a photosynthetic spike - in most other trichome-forming plants (like tomatoes) this means more essential oil production.
The power drivers might fuck off and die, yes, and occasionally a poorly-binned diode might fuck up a whole panel if not properly designed, but what you talk about, over the 4+ year current life of the majority of my LED units, only two out of several THOUSAND have failed.
"You could achieve a better result with a mix of cool and warm fluorescent lights or daylight spectrum fluorescent."
No, you can't. Fluorescent lighting is nowhere near as efficient, and LED outputs much higher photon flux densities per watt, unfocused, than an equally-powered and focused fluoro tube, with the option of selecting more efficient wavelengths (white fluorescents use tri-phosphor tech. White LEDs now have more broad peaks, providing near-incandescent quality light at well over HID efficiencies.)
We've got LED kicking the pants off of CFL by roughly triple the photon flux density per watt.
Wikipedia is not reliable. Those sources are way outdated. We've got 1W diodes pumping 150 lux/W and Cree has already smashed past 200 lux/w and did that last year. Arrays of these diodes pump out huge amounts of light.
~designs horticultural LED and interior LED lighting
Yes, it's called the dependencies issue that Java has. You don't get that with ASM as you've coded all of your own stuff. The Java devs failed to realize that.
" This destroys the profitability of the project for the customer "
Not if you are a programmer worth anything, no.
"Java lets you be a lazy and wasteful dev and still be productive."
If you call releasing something ridden with security holes and poorly-optimized productive, then yes.
And I have tons of experience working with large scale development. Starting from Ingram Micro to Hewlett-Packard to global horticultural companies across the world.
I've had entire government IT departments (through their agricultural branches) under my control. You know what I did? Got them the hell out of my way when I saw how crappy their desired programming choices were, and how poorly their code ran on my dev systems.
" Well, I could be the same and wail about how you almost certainly don't understand the physics behind the circuitry "
Except you'd be dead wrong there, as not only do I know that, but I also know how to bypass the need for quantum particles such as photons, and directly stimulate biological systems with direct energy impulses, creating the first zero-light horticultural technology. I've built it all from simple LED lights to millions-of-units-monthly production lines, from scratch.
"It is fairly clear you wish to be an elitist."
Not wish, already am. You depend upon me to keep your food cheap, and you don't even realize it.
"knowledge at that level is simply not needed anymore to be productive"
Just wait until all the ASM programmers die off and you guys want to introduce new hardware - have fun with that.
" It is simply not possible for a single individual to construct everything on a huge game by themselves within commercial timespans"
Richard Garriott would be more than happy to prove to you just how wrong you are.
"a lack of experience with non-Intel hardware"
DEC Alpha to ARM to Cell to even the venerable TMS9900, and the TMS9900 was when I was 5 - try again.
"Again, your perspective of someone who plays with tiny toy demos shows through"
25 years of experience sees right through your nonsense. Some of those exact 'toy demos' have their software used in many games and other applications, like the voice synthesizer in the 64k demo candytron, which is used in my voice notification system in my horticultural facilities, or the animation software, which happens to be linked to a remote robot that I use to 'run' through my UK research lab and send me information back here in California so I don't need to get on a plane every other week.
"I just don't use it because it won't help me get the job done in minimal development time"
That very laziness is why the software you develop will have bugs, get cracked, and pirated.
" Futhermore, the time required to develop procedural textures in a large game is not worth it."
The brick texture took a mere few seconds to make. I see you've never bothered to get and work with the tools they used to make the game so you're just shooting blind from your mouth. And those tools are freely available with full documentation.
"I have single textures that are thousands of times bigger. Increasing my development time by using ASM so I can create a tiny 64 KB program is completely meaningless when modern hardware is multi-gigabyte and the game resources I have are similarly sized. "
Enjoy your easily-cracked and poorly-optimized code. What should run on a Pentium 3 with 128MB of RAM now requires a quad-core and 16GB of RAM. Is that supposed to impress me?
"Don't confuse this situation with the world outside yours"
Considering you fail to see the direct link between both worlds, I am not sure you're even logically able to make that argument.
"What you need to understand here is that data stored "in the cloud" is data stored in leased property. That is, you store the data in property owned by someone else who has conferred to you access rights to use their property for storage--in fact, Web services like AWS hosted servers could be considered similar to living and operating space. "
Then you better learn what the fuck Right to Possession means, if you're going to use such a flawed logical analogy, and see why your words just went to shit with that supposition.
"Well it turns out for huge problems a single person is simply not productive enough to complete the task."
Thirty Java programmers failed to do in one year what I did in two months for the horticulture industry.
" It turns out I've worked on many, many development project as a consulting developer and architect."
So, an outsider, and not directly and intimately involved with any of the industries you work in. Gotcha.
"However, this topic was about gaming and I'm very interested to hear your experience in game development and why you think the choice of Java for my combat flight simulator in Java is a mistake (note also that the wildly popular Minecraft is written in Java, and Bohemia Interactive's Arma3 and Take On Helicopters have Java APIs)."
That's an easy one - Have you seen how buggy and poorly-optimized Minecraft and Arma are? Maybe you should go to the forums and listen to the people bitching. I just so happen to be one of them, having owned both games.
ASM for all components that require rapid execution capability and speed, not security-ridden JIT. Also, raw ASM makes it MUCH harder to crack, should DRM be implemented/written in ASM, as you need INTIMATE knowledge of how the programmer works and codes to even be able to get close to figuring out the obfuscated code that most ASM compilers will produce.
"Indicidentally, I've done a *lot* of device control work. I used to have to use C and C++ for memory constrained devices a decade and a half ago."
Again, ASM wins for memory-lacking applications and hardware. That's why it exists, as native machine code that can be run realtime without having to wait on much of anything else. Ever play with MenuetOS? Do it sometime, and watch games ported to raw ASM for it (like quake) run in a software mode at high resolutions, no 3-d rendering, and still look almost as good as hardware-accelerated version.
ASM forces you to code properly. Java lets you be a lazy and wasteful dev. ASM teaches you how hardware will work and its intricacies, Java teaches you how hardware will fail because of poor JIT execution and poor branch prediction.
" As far as I can see Java memory usage is on par with other big games"
Oh? Comparing Q3 to QuakeLive (which uses Java) I get roughly 1/50th the performance using QuakeLive, that's in both loading times and overall FPS.
"So I'd be interested to hear if there was a factor I had missed and a reason why you recommend ASM for this multi-platform project."
Go look up the Demoscene and learn what *REAL* programmers can do in a mere 96KB of code using pretty much solid ASM -.kkreiger is the immediate example I'd point out for a demo game, then I'd say go check out the 64K demos. Your game could very likely have been built much faster and made much smaller and perform better. C/C+/C#/Java simply do not, can not, and will never compare. These demoscene guys have tried. It just can't happen.
So - Security, Speed, Satisfaction, Safety, Assurance of Quality Code (because it won't run otherwise!) and much more are reasons you should be learning ASM. On top of that, since most hardware out there is of the x86 variety (exception being gaming consoles) your x86 ASM code should run ANYWHERE if you can actually program every bit yourself, it won't matter the OS, it won't matter the API, as your game has all it needs baked in and can simply avoid all the overhead cruft of the OS, thus gaining far more performance (assuming you can actually program the graphics engine to handle both VLIW and GCN for AMD/nVidia cards, which will be your biggest problem.)
Slashdot Mirror
I would be willing to bet the 60Wh quote is in regards to the monitoring system and pumps needed to move the nutrient solution, not to move the towers.
"They can't be close together, or they'll be in each other's shade."
That problem got solved long ago. Rotating floors. Allows for much more even sunlight dispersion across the entire crop.
"You aren't going to get one through hydroponic farming. Hydroponic tomatoes look beautiful and big but they are dense and flavorless."
That would be the cultivar chosen for mechanical picking and shipment, plus a combination of poor nutrient choice. Get an heirloom cultivar and you will notice a world of difference in taste, using the same nutrients. Using more diverse stuff like SEA-90 (supplementing for the nitrogen and potassium [I think, don't have that chart handy right this second]) will show even better results.
Here's a soil mother (two months old) tomato and basil versus a hydroponics tomato and basil clone (three weeks old.) Look at the difference.
The basil alone was DRIPPING with pungent oils, and much larger leaves. I had pesto that had my local Italian neighbors raving for WEEKS.
"This is probably because they are not able to farm the bees needed for tomato pollination."
Most tomatoes commercially produced are self-pollinating and have been for well over 50 years.
It's rockwool NFT.
If you have a good room with barriers, not much pest control.
That's not vertical farming, though. That's rotary. Vertical implies large stacks of trays or channels. Most weed growers use 'vertical' for the vertically-oriented NFT systems, but that's not true vertical.
They taste fine. I've done a grafting of potato rootstock + Medusa peppers. Potatoes weren't spicy, peppers weren't starchy and properly spicy (though it looked really strange having a part potato vine loaded with clusters of peppers.)
"I hope they consulted with the experts in the field, the weed growers."
Those aren't experts. Expert botanists use photon flux density as their measurement, not lumens. Any electrician will tell you how bullshit the term "grams per watt" is when used how the weed growers use it. Any photobiologist will know that going past ~3,000 umol (50% more photon flux than the sun puts out per meter at sea level mid-summer at noon) is a BAD idea for most plants, including cannabis (yet idiot growers want those white, photosynthetic-bleached 'frosty' nugs.)
The fact that you refer to these people as 'experts' is laughable and shameful to those of us that have REAL expertise in this field, across more crops than just petty weed and tomatoes and lettuce.
Shit I can get you 150+ lumens per watt but that isn't what matters, photon flux density per wavelength is what matters.
And I can custom-build you a panel at pretty much near-cost. I've got the capability to manufacture units so small and so powerful that Intel would be scratching their heads wondering how the fuck I keep them cool.
500w in 30mm x 30mm? Easily done for the raw array. Copper-backed, attached to a large aluminum radiator, no problem. Power supply that's reliable? You want that internal or external?
It's only expensive because you're looking at scamming resellers of Chinese stuff. Go check out Hydrogrowled, then look at the patents they claim, only to realize that's what every other Chinese manufacturer on the planet has already claimed and made, well before those patents were even filed.
"There's a narrow absorption spectrum of chlorophyll A and B (P680,P700). Sunlight is mostly wasted on plants."
Whomever modded this up is wrong wrong wrong just like the poster.
In fact, green light over ~400-450 umol is more efficient in terms of quantum yield.
Almost all wavelength ranges have some effect on plant physiology and development, this includes UV and IR and green.
Source: I am a horticultural researcher for a global company. Specifically, I design and test their wavelength-and-crop-specific lighting.
"Everyone keeps mentioning LED but LED is insanely expensive lighting"
No it's not. You can order direct and custom from the Chinese, 90w panels to your specs for like $75 each - the REAL issue is whether or not you know what you WANT in the panel. That might cost you extra (Like aluminium thermal PCB instead of PolyPCB+Heat Plate attachment) and such.
Cut out the middleman and go direct to the REAL manufacturer.
Let me guess, you looked at the scammer site hydrogrowled.com, didn't you?
"Maximum theoretical overall efficiency of LED lighting is only 43.9 % (http://en.wikipedia.org/wiki/Luminous_efficacy)."
Outdated information. Did you forget the LED in the lab that output more light energy than it consumed, by virtue of the surrounding thermal energy? That was a story right here on slashdot not even a month or two ago.
Also, 299 lumens per watt for LED? Cree's already hit over 220 in the labs. We have new topology and substrate manufacturing that eliminates many of the inefficiencies of LED (backscatter, re-absorption of emitted photons,) that have yet to be implemented in common process.
At the rate I've watched LED tech progress, in maybe three years we'll hit that barrier and break it, we have power supplies with power factors of 0.98, the only inefficiency is A. source power and B. LED
"I understand plant photosynthesis can't use the IR, a narrow green band, and UV portions of sunlight"
This is incorrect. IR has its uses in triggering flowering hormones plus day/night cycles, plus it has an additional photosynthetic effect. Green is used - its quantum yield is higher than blue or red, but ONLY when the photon flux density gets roughly over ~400 umol (which most vegetative/non-fruiting crops do well within 150-300 umol) and UV has photosynthetic and carotenoid reactions - and for those growing cannabis, UVB light specifically has a boosting effect upon THC production in plant tissues and UVA has a photosynthetic spike - in most other trichome-forming plants (like tomatoes) this means more essential oil production.
A few underground cistern tanks would be more than enough. The issue would be water treatment.
That would depend upon whether or not they did just the roof or the entire building with solar cells.
Just the roof, no. The entire building, yes.
"LED's give a high electrical efficiency but they aren't cost efficient even with the electrical savings"
Absolute nonsense. Several times I've had LED payback in terms of months, not years, over several intense-light required crops.
"he LED's last forever thing is a myth especially given the larger number of elements to fail."
I'm holding a Nichia LED right now. Just the bare LED on a tiny square thermal board. I smashed it face-first with a hammer, exposed the substrate and surface topology...
As you can see, it still fucking works.
The power drivers might fuck off and die, yes, and occasionally a poorly-binned diode might fuck up a whole panel if not properly designed, but what you talk about, over the 4+ year current life of the majority of my LED units, only two out of several THOUSAND have failed.
"You could achieve a better result with a mix of cool and warm fluorescent lights or daylight spectrum fluorescent."
No, you can't. Fluorescent lighting is nowhere near as efficient, and LED outputs much higher photon flux densities per watt, unfocused, than an equally-powered and focused fluoro tube, with the option of selecting more efficient wavelengths (white fluorescents use tri-phosphor tech. White LEDs now have more broad peaks, providing near-incandescent quality light at well over HID efficiencies.)
Sorry, you're not qualified to be discussing this. I happen to do this globally. (that's my UK research facility with me doing the testing.)
Go look all over alibaba.com - most of the newer 1000w spotlights are pushing 170,000 lumens or more, for 170l/w.
It's already been mainstream for well over a year.
I refuse to update that place. They'll as always cite 'original research' and wipe it out. Had it happen many times.
Learn how tenant Right to Possession (arguably the person leasing datacenter space is a tenant) affects your RRights as shown in the Bill of Rights.
Quite clearly, 4th amendment protections apply.
Was that really such a hard logical conclusion to make with maybe one or two google searches?
No, they are not.
We've got LED kicking the pants off of CFL by roughly triple the photon flux density per watt.
Wikipedia is not reliable. Those sources are way outdated. We've got 1W diodes pumping 150 lux/W and Cree has already smashed past 200 lux/w and did that last year. Arrays of these diodes pump out huge amounts of light.
~designs horticultural LED and interior LED lighting
"There must have been other factors in play. "
Yes, it's called the dependencies issue that Java has. You don't get that with ASM as you've coded all of your own stuff. The Java devs failed to realize that.
" This destroys the profitability of the project for the customer "
Not if you are a programmer worth anything, no.
"Java lets you be a lazy and wasteful dev and still be productive."
If you call releasing something ridden with security holes and poorly-optimized productive, then yes.
And I have tons of experience working with large scale development. Starting from Ingram Micro to Hewlett-Packard to global horticultural companies across the world.
I've had entire government IT departments (through their agricultural branches) under my control. You know what I did? Got them the hell out of my way when I saw how crappy their desired programming choices were, and how poorly their code ran on my dev systems.
" Well, I could be the same and wail about how you almost certainly don't understand the physics behind the circuitry "
Except you'd be dead wrong there, as not only do I know that, but I also know how to bypass the need for quantum particles such as photons, and directly stimulate biological systems with direct energy impulses, creating the first zero-light horticultural technology. I've built it all from simple LED lights to millions-of-units-monthly production lines, from scratch.
"It is fairly clear you wish to be an elitist."
Not wish, already am. You depend upon me to keep your food cheap, and you don't even realize it.
"knowledge at that level is simply not needed anymore to be productive"
Just wait until all the ASM programmers die off and you guys want to introduce new hardware - have fun with that.
" It is simply not possible for a single individual to construct everything on a huge game by themselves within commercial timespans"
Richard Garriott would be more than happy to prove to you just how wrong you are.
"a lack of experience with non-Intel hardware"
DEC Alpha to ARM to Cell to even the venerable TMS9900, and the TMS9900 was when I was 5 - try again.
"Again, your perspective of someone who plays with tiny toy demos shows through"
25 years of experience sees right through your nonsense. Some of those exact 'toy demos' have their software used in many games and other applications, like the voice synthesizer in the 64k demo candytron, which is used in my voice notification system in my horticultural facilities, or the animation software, which happens to be linked to a remote robot that I use to 'run' through my UK research lab and send me information back here in California so I don't need to get on a plane every other week.
"I just don't use it because it won't help me get the job done in minimal development time"
That very laziness is why the software you develop will have bugs, get cracked, and pirated.
" Futhermore, the time required to develop procedural textures in a large game is not worth it."
The brick texture took a mere few seconds to make. I see you've never bothered to get and work with the tools they used to make the game so you're just shooting blind from your mouth. And those tools are freely available with full documentation.
"I have single textures that are thousands of times bigger. Increasing my development time by using ASM so I can create a tiny 64 KB program is completely meaningless when modern hardware is multi-gigabyte and the game resources I have are similarly sized. "
Enjoy your easily-cracked and poorly-optimized code. What should run on a Pentium 3 with 128MB of RAM now requires a quad-core and 16GB of RAM. Is that supposed to impress me?
"Don't confuse this situation with the world outside yours"
Considering you fail to see the direct link between both worlds, I am not sure you're even logically able to make that argument.
"What you need to understand here is that data stored "in the cloud" is data stored in leased property. That is, you store the data in property owned by someone else who has conferred to you access rights to use their property for storage--in fact, Web services like AWS hosted servers could be considered similar to living and operating space. "
Then you better learn what the fuck Right to Possession means, if you're going to use such a flawed logical analogy, and see why your words just went to shit with that supposition.
Let furries rape them.
I think they'll be begging for death after that.
"How can you spoof the caller ID anyway?"
Relatively easy. Yesterday I had . - yes a period - show up as the caller ID number.
I don't own dogs.
"Well it turns out for huge problems a single person is simply not productive enough to complete the task."
Thirty Java programmers failed to do in one year what I did in two months for the horticulture industry.
" It turns out I've worked on many, many development project as a consulting developer and architect."
So, an outsider, and not directly and intimately involved with any of the industries you work in. Gotcha.
"However, this topic was about gaming and I'm very interested to hear your experience in game development and why you think the choice of Java for my combat flight simulator in Java is a mistake (note also that the wildly popular Minecraft is written in Java, and Bohemia Interactive's Arma3 and Take On Helicopters have Java APIs)."
That's an easy one - Have you seen how buggy and poorly-optimized Minecraft and Arma are? Maybe you should go to the forums and listen to the people bitching. I just so happen to be one of them, having owned both games.
ASM for all components that require rapid execution capability and speed, not security-ridden JIT. Also, raw ASM makes it MUCH harder to crack, should DRM be implemented/written in ASM, as you need INTIMATE knowledge of how the programmer works and codes to even be able to get close to figuring out the obfuscated code that most ASM compilers will produce.
"Indicidentally, I've done a *lot* of device control work. I used to have to use C and C++ for memory constrained devices a decade and a half ago."
Again, ASM wins for memory-lacking applications and hardware. That's why it exists, as native machine code that can be run realtime without having to wait on much of anything else. Ever play with MenuetOS? Do it sometime, and watch games ported to raw ASM for it (like quake) run in a software mode at high resolutions, no 3-d rendering, and still look almost as good as hardware-accelerated version.
ASM forces you to code properly. Java lets you be a lazy and wasteful dev. ASM teaches you how hardware will work and its intricacies, Java teaches you how hardware will fail because of poor JIT execution and poor branch prediction.
" As far as I can see Java memory usage is on par with other big games"
Oh? Comparing Q3 to QuakeLive (which uses Java) I get roughly 1/50th the performance using QuakeLive, that's in both loading times and overall FPS.
"So I'd be interested to hear if there was a factor I had missed and a reason why you recommend ASM for this multi-platform project."
Go look up the Demoscene and learn what *REAL* programmers can do in a mere 96KB of code using pretty much solid ASM - .kkreiger is the immediate example I'd point out for a demo game, then I'd say go check out the 64K demos. Your game could very likely have been built much faster and made much smaller and perform better. C/C+/C#/Java simply do not, can not, and will never compare. These demoscene guys have tried. It just can't happen.
So - Security, Speed, Satisfaction, Safety, Assurance of Quality Code (because it won't run otherwise!) and much more are reasons you should be learning ASM. On top of that, since most hardware out there is of the x86 variety (exception being gaming consoles) your x86 ASM code should run ANYWHERE if you can actually program every bit yourself, it won't matter the OS, it won't matter the API, as your game has all it needs baked in and can simply avoid all the overhead cruft of the OS, thus gaining far more performance (assuming you can actually program the graphics engine to handle both VLIW and GCN for AMD/nVidia cards, which will be your biggest problem.)